Non-operative treatment of knee osteoarthritis has centered about the relief of pain accomplished primarily by the use of analgesic and/or anti-inflammatory agents. Although providing short term benefit, these treatments have been reported to accelerate the disease process and the long-term consequences of their use is unknown. Little or no objective data exist regarding the biomechanical or biochemical consequences of these therapies in man. This proposal will test the hypotheses that (1) non-operative treatment (oral analgesic, oral non-steroidal anti- inflammatory, and intra-articular steroid) or patients with painful knee OA will result in gait alterations, specifically an increase in the external load, and (2) there will be no significant differences among the treatment groups. Correlation of gait changes with clinical symptoms (pain) and with biochemical parameters (cartilage matrix markers) will also be undertaken. Patients entered into the study must have idiopathic knee OA with medial compartment involvement assessed as radiographic stage II or III. Patients who flare after wash-out from their current medications will be randomized in a double-blind manner to one of the 3 treatment groups or placebo, evaluated clinically, and undergo an initial gait analysis. Patients will return for repeat clinical and gait studies after 2 weeks of treatment and then at 3 months post entry. Joint aspiration and injection will be done on all patients after the first gait study and repeat aspiration at the end of the study to obtain synovial fluid for marker determinations.